Revision History
GGUF · 14 revisions
Sizes are character counts of the article source. The signed number is the change from the previous revision.
Recent edit summaries
Detailed summaries recorded by editors. Generic maintenance summaries are omitted here; every recorded revision remains below. Dates describe the edit, not necessarily the event it covers.
- Correction: Corrected the 7B file-size table (Q8_0 is about 7.2 GB, not 13-14 GB), an unsourced Llama 2 perplexity series, a misreading of Justine Tunney's mmap post (she found no benefit from madvise), the false claim that every gguf-split shard keeps full metadata, the stale 50 GB Hugging Face file limit and 45,000 GGUF-checkpoint count, the claim that gpt-oss was trained natively in MXFP4, and an unsupported description of Unsloth Dynamic 2.0 (plus 21 minor fixes)
Version 14 · Nov 2, 2026, 10:22 AM
- Update: added the Q1_0 (1-bit) and Q2_0 (2-bit, used for ternary weights) types merged into llama.cpp in April and July 2026 for PrismML's Bonsai models, with citations to the pull requests.
Version 13 · Oct 30, 2026, 10:55 AM
- Added a related article link
Version 12 · Oct 21, 2026, 01:33 AM
- Improve internal links (2026-10-13)
Version 11 · Oct 13, 2026, 10:09 AM
- Added 6 contextual internal links
Version 10 · Jul 23, 2026, 03:05 PM